12/20: Released a Basel Face Model to FLAME converter.12/20: Released DECA, a framework to reconstruct FLAME meshes with animatable details from images.02/21: FLAME vertices masks available for download.04/22: Released EMOCA, a framework to reconstruct expressive FLAME meshes from images.07/22: Released MICA, a framework to reconstruct metrical 3D face shapes from images.07/22: Released MICA dataset obtained by unifying existing 3D face datasets under a common FLAME topology.11/22: Collection of public FLAME resources (i.e., code, publications, data).05/23: Updated FLAME with revised eye region (FLAME2023).FLAME is significantly more accurate and is available for research purposes. We compare FLAME to these models by fitting them to static 3D scans and 4D sequences using the same optimization method. FLAME is low-dimensional but more expressive than the FaceWarehouse model and the Basel Face Model. In total the model is trained from over 33, 000 scans. We accurately register a template mesh to the scan sequences and make the D3DFACS registrations available for research purposes. The pose and expression dependent articulations are learned from 4D face sequences in the D3DFACS dataset along with additional 4D sequences. FLAME combines this linear shape space with an articulated jaw, neck, and eyeballs, pose-dependent corrective blendshapes, and additional global expression blendshapes. FLAME uses a linear shape space trained from 3800 scans of human heads. Our FLAME model (Faces Learned with an Articulated Model and Expressions) is designed to work with existing graphics software and be easy to fit to data. We seek a middle ground by learning a facial model from thousands of accurately aligned 3D scans. At the low end, face capture from consumer depth sensors relies on 3D face models that are not expressive enough to capture the variability in natural facial shape and expression. At the high end, the best facial animation is indistinguishablefrom real humans, but this comes at the cost of extensive manual labor.
The field of 3D face modeling has a large gap between high-end and low-end methods.
0 kommentar(er)
